How to make a risk-of-bias plot
- Choose the assessment tool you used: RoB 2, ROBINS-I, QUADAS-2, RoB 1 or another tool.
- Paste one row per study from Excel, Google Sheets or your screening software, or drop an .xlsx or CSV file. The Excel template for each tool has the right columns.
- Check the reading under the data box: which column holds the study names, which are domains, and whether Overall and Weight were found. Judgements that are not recognised are listed above the figure.
- Choose the plot: traffic light, summary, or both in one figure, and whether the summary is weighted.
- Apply your journal’s rules and export a vector PDF or SVG, or PNG/TIFF up to 1200 dpi. The methods text and a robvis script are one click away.
Reporting risk of bias
- PRISMA 2020 asks for the methods used to assess risk of bias (item 11) and the results for each study (item 18); a traffic-light plot covers item 18 compactly. Pair it with a PRISMA flow diagram.
- Report the overall judgement per study and how it was derived; with RoB 2, a study is at high risk overall if any domain is at high risk, or if some concerns in several domains substantially lower confidence in the result.
- Use the assessment for each outcome that you meta-analyse, and consider a sensitivity analysis without high-risk studies in your forest plot.
Data format
| Study | D1 | D2 | D3 | D4 | D5 | Overall | Weight |
|---|---|---|---|---|---|---|---|
| Adams 2015 | Low | Low | Some concerns | Low | Low | Some concerns | 14.2 |
Frequently asked questions
Which risk-of-bias tools are supported?
RoB 2 for randomised trials (5 domains), ROBINS-I for non-randomised studies of interventions (7 domains), QUADAS-2 for diagnostic accuracy studies (4 risk-of-bias domains) and the original Cochrane RoB 1 tool (7 domains). For other tools, such as ROBINS-E or ROBINS-I version 2, choose “Other tool”: the domain names are taken from your column headers.
What should my table look like?
One row per study: the study name, one column per domain (named D1, D2 … or with the full domain names), then an optional Overall column and an optional Weight column. Judgements can be written out (Low, Some concerns, Moderate, Serious, High, Critical, Unclear, No information) or abbreviated (L, SC, H, +, ?, X, NI), so exports from Covidence, RevMan or Excel usually paste as they are.
What is the difference between the traffic-light and the summary plot?
The traffic-light plot shows every judgement for every study, so readers can see which study has which problem. The summary plot shows, for each domain, the proportion of studies (or of the meta-analysis weight) at each level of risk. Cochrane reviews usually include both; many journals accept either.
Should the summary plot be weighted?
Weighting by the meta-analysis weights shows how much of the evidence, rather than how many studies, is at each level of risk, which is what matters for the pooled result. robvis supports both and weights by default. Add a Weight column (for example, the percentage weights from the forest plot) and tick “Weight the summary”.
Are the colors colorblind-safe?
The default palette is a traffic light built from the Okabe–Ito colors, which stay distinguishable for the common color-vision deficiencies. Every circle also carries a symbol (+ low, − some concerns or moderate, ? unclear or no information, × high or serious, ! critical), so the plot reads in grayscale too. The Cochrane/robvis colors are available if your review group expects them.
Can I reproduce the plots with robvis?
Yes. “R code” gives a robvis script with your data, the tool and the options you chose. robvis draws with its own sizes and colors; the CleanFigure figure is sized in millimetres and points for your journal.